transcript

Can you please introduce yourself, and what you do? What's your role?

So, my name is Ashraf Rahman. My role is a Solutions Engineer in Eigen Technologies. So that basically means I work with AI, and implement AI in different businesses based on what they need, right? So our main product is to work with documents and extracting as much information as we can from that. So we can give our clients the necessary information they need and make it a lot more presentable, rather than them sifting through hundreds of documents to extract the same level of information. So we primarily use natural language processing for that. Natural language processing, or NLP, is a subsector of deep learning. So, that's how everything comes together and that's why we use AI to extract those documents.

That leads me on to my next question. For those who don't know, what is ‘deep learning’ and what is ‘machine learning’? And how do they differ? And how do they relate to artificial intelligence?

So deep learning is basically a sub-sector of machine learning, right? So machine learning, it's more with like, pattern recognition. So a couple of things would be like if you wanted to recommend a product to, let's say, a customer right, you would use machine learning most likely. Because what that would do, is it would understand what the customer is already buying, and then recommend something along those lines.

So that's one area. But in machine learning, there's actually like a couple of categories, which is supervised learning and then unsupervised learning. Supervised learning is basically you teach it everything you can around how to identify one specific pattern and you just train it to identify that, right? So, if you knew nothing about like project management, but I gave you all the resources and I showed you everything to do with project management to identify, let's say, if a project's going to go past the deadline. If I'm giving you everything and I've trained you for enough time, then you're going to understand it, you're going to identify if a project is going to go past this deadline or not, right?

So with unsupervised learning it's basically like, ‘here's what I need, and go find something around that area’. It's a bit more- it has its pros and cons. I can go in for like ages on it, honestly. But the best overview I can give you is, let's say, again with recommending, let's say a recommender system, right? You have a recommender system. And in the recommender system, you have something that recommends products to people. Now you want to group those people into specific categories, so then you can recommend even more products to them. So then, what unsupervised learning would do, is categorize those people based on where they are numerically on a graph, and then put them into a group. And then you can go in deeper and like add more recommender systems onto it to recommend more products, based on what this specific group likes. So that's unsupervised learning and yeah, based on what the customer buys, and likes, and doesn't like, they can move around in this little graph, and then that category changes for them. Sort of like TikTok, when you only watch, let's say memes, or if you only watch STEM videos, then you're going to get recommended STEM videos. Right? Because that's what the system recognizes, and groups you into that category basically.

So the difference between supervised and unsupervised learning, is how the artificial intelligence sources its data?

It’s how involved you are with it basically.

Yeah.

Unsupervised learning is basically like ‘here's all the things that we need, go and do it’. Supervised learning is ‘this is what I want you to find, this is the data I've trained you on’, I'm holding its hand basically.

So with the deep learning, it's more like neural network based, right? So, it's about mimicking- it's another form of pattern recognition, I would say, in my opinion, but it's more linked to neural networks. So pattern recognition, but in a sense of how the human would recognize the pattern, right? And because of that, again, it goes through layers, right? Just like our brain has like synapses, the neural network has layers. If we were given a picture of something that looks like a penguin, we'd be able to tell that it's a penguin because we'd look at like different features and different parts of the picture; to be like, okay, this is probably most likely a penguin because it has an orange beak, or white belly, things like that. But yeah, with deep learning as well, it has its own set of categories. So you have like image recognition, natural language, speech recognition, video recognition, and robotics. So again, the way that they work is, it's a much more deeper level of understanding for the machine, in a sense, where they are able to sort of adjust their own parameters. And that's what sets it aside from machine learning. With machine learning, we adjust its parameters, right? So we do the research and we figure out, okay, it does the best learning with X or Y, so we're going to set those parameters in and let it learn based on the restrictions we put in. With neural networks, it's sort of, ‘this is what I need, and I want you to identify something based on what I give you’. And then basically what it would do is it would adjust itself, based on how many layers it has. Remember, with layers, we talked about synapses in the brain. So the more layers we have, the better is to solve the problem, essentially. And yeah, so that’s the biggest difference.

OK, nice, thank you. If you had to evaluate the differences between an artificial intelligence and a human completing a specific task, what would be the key differences or similarities that you might see?

So the first would be speed. The speed on the AI, is just hugely efficient compared to a human brain, just because of what's behind it. So let's say, if we put Chat-GPT on a Mac, right, it's probably going to take like days or months, for it to answer the question compared to like a quantum- something like heavy duty. Obviously with speed, AI wins. With accuracy, AI wins as well, because it's connected to a substantial amount of information, than a human is. If we get asked the question, and AI gets asked the question, they already have a database that they can go and reference to, but we have to do it within our memory, right? Unless you also give us a database that we can access with just the same amount of speed as they can, which is not -

Physically impossible.

Yeah, exactly. Like precision as well. Sometimes, because we don't have access to that information as fast as they do and immediately, we're less precise compared to them. Where, they basically have all the information they need to answer your question. Those are the three things. But when it comes to things like chat-bot services, where humans are interacting with AI services, there is a clear - well, not clear, I would say that that border is starting to become thinner and thinner by the day; but before, there was a clear like indication whether or not it was a bot, essentially.

Yeah. And it's harder to distinguish and define artificial intelligence.

Yeah. So I would say now, if it can answer your question really fast and on point, then it's probably AI. And because it's studied the way that human behaves, and interacts with each other, it sort of mimics that interaction between itself and you.

It’s a very loose definition as well. Artificial intelligence, it's “officially” defined in some cases, as just a theory or software that can do what a human can. And like you were saying, do you think that that poses a threat? That we can't define artificial intelligence as easily as we perhaps once could?

I would say- to me, I think not now. I don't think it's a threat now. Unless with Boston Dynamics, what they're doing building, walking robots and stuff. I think when they manage to put Chat-GPT in a Boston Dynamic robot, then I'd be worried. But in the meantime, it's sort of like the 80’s with the Excel, with Microsoft Excel, right? All the accountants were worried because it’s going to take their job, it’s doing formulas faster than they can. But then what ended up happening is now it's become a tool, rather than a thing that can replace your job, right? But here's the thing, if you don't use that tool, you're going to get left behind. Like what would you rather have, would you rather have an accountant that uses Excel, or does it by hand on a calculator? It just doesn't make sense. And I think now we're getting into that point where you're better off learning how to use AI for your advantage, than try to fight it. Because it's not something that you can fight. Even, I’ve known a couple of startup companies, actually, that are trying to incorporate Chat-GPT to help their developers out. Because, yeah, I would say the key element that Chat-GPT, or AI, is missing, is creativity. It can mimic creativity, but it can't be creative. You know what I mean? You have to tell it what you need. So in a sense, you still have to be creative in order to use AI to build what you want to build. It just takes less effort.

Do you think that in the future it could have that creative input? It could, I don't know, conduct that creative part of that process?

I think- I think maybe. But we’re looking like-

A long way away?

Yeah, I want to say, like, between maybe 5-10 years. But if then, even that, it can just give suggestions. And I think that even if AI takes over the creative space, what's going to be more valuable at the end of the day is, if a human does it. Because it's going to be so common to see AI doing it, that the value of human art is going to start to increase. That's what I think would be- because it will be limited. They'll be less human art. So therefore, it’s a valuable resource.

Yeah, I was talking to someone about it and we were talking about how it feels like on the surface, it looks like it's democratizing art, and making it more accessible for a lot of people. But that eventually, if AI does - and likely will - become this huge thing that's widely used, and widely known. All it's going to do is begin feeding off of each other. Because all of that database is going to be pulled from an AI five years ago, that's going to be pulled from an AI two years before that, and it's going to become this, sort of incestuous mound. But like you're saying, that could place higher value on human creativity and human autonomy in work, or in art, and any different area.

Because think about like this, right? So, Bing is implementing the chat-bot onto their search engine, right? And rather than doing what Google is doing by providing the links when someone searches for something, it does the link extraction and just formulates an answer for the user, right; but, if nobody is giving, like, a human input and it's all done by AI, at the end of it it's just going to become a whole mess because it needs human input to provide information for it to use. So, unless it can do something where it's able to generate content and gather content simultaneously, then there's still that need for, you know, ‘human in the loop’ we call it.

How do you feel about the recent increasing “hype” around AI software such as Chat-GPT, or Midjourney, Dall.E, Stable Diffusion, and all of the big names that are out there at the moment, they're getting a lot of attention. How do you feel about that in general?

Yeah, I think it's great. It’s paving the path of what will soon be common in the future. And it’s almost like it’s inevitable either way. Technology is constantly changing and improving. As a software engineer, I have to keep up with new emerging technologies and see how we can use it to improve our work. Right? The key is to see how we can use it, right? And if it makes people's lives easier, then people are going to start using it. So there's definitely a need for it. And I think it's just about how you can use it, rather than just fight it, because then there's no fighting it.

We've opened that box now, it's happening.

Here's the thing, if we fought the moon landing, let's say, we wouldn’t have half of the stuff that we have now, because we’ve built on the research and discoveries scientists made on building the rocket, to what we have now. It's the same thing, this AI looks scary and everything, but once you realize how to use it to your advantage, it can take you to different places. There's job roles now, they're popping up, called ‘prompt engineers’, and I think that's great. It's like they’re hiring people to basically create prompts and utilize Chat-GPT in like different ways - rather than how a typical user would use it - to just get much more specific and detailed answers. And yeah, that's just an example of something that's completely new that's coming.

We've kind of touched on this already, but do you feel like people have a right to be worried about potentially losing their job or their income?

I would say, yes. For some, yes; for others, I would say it would be, if you don't use it. But again, we can compare that to telephone operators. Right? We don't need telephone operators right now, because technology has changed, but now we have more jobs that that have come from it, right? Call service, and all those other different types of- telemarketers, for example. Same thing with AI, I think. Yeah, it's going to close a lot of jobs, it's going to open some jobs as well, right?

Adapt or die.

Yeah.

Yeah.

Another point we’ve kind of already brushed over a little bit. How do you feel about artists claims towards organizations like Midjourney, or similar, in terms of that creating infringing content, or using their artwork in in their databases without permission?

That's the thing, if you say that, that means nobody else can paint what Van Gogh painted, or nobody else can paint what Da Vinci painted. You're not stealing the work, you're just mimicking it, right? It's not- I don't think it's that. I think it's more like an idea that, okay, ‘that's a great idea use some of that; oh, that's another great idea, use some of that’. That's how Midjourney can make an image that's, let's say, a mixture between like so many different artists.

Yeah, that's been the root of art, since the beginning of art. It can only ever be inspired by the people who have come before you. It is already - like the word we used - it’s already pretty incestuous. That is art, that's always been art. I think it's impossible to be original.

But saying that, if it’s exactly what an artist has done, then copyright.

That’s different, yeah 100%. That would be their intellectual property, obviously depending.

But I think part of it as well is coming from rather than a fear of their work being stolen; but I think there's a general fear, or sense of anxiety, around a lack of autonomy. This idea that potentially in artificial intelligence - I don't want to use the word ‘uprising’ because it feels very iRobot - but the idea that artificial intelligence could have a level of control or dictatorship over the discourse of human history in the future. Do you think that artificial intelligence will have a significant impact on our future?

Yeah, 100%. Like look at self-driving cars, right? That's going to be such a big thing. Uber actually, they don't really own any vehicles, right? But Uber is actually working on driverless cars. And with driverless cars being involved, that means a lot of people are going to be ordering taxis with no drivers in it. That means people's lives are going to be changed. Yes, there'll be less taxi drivers, but there'll also be more comfort when you drive. See how there's like two sides of everything.

I think that, yeah, it is going to change our lives because look at Alexa, look at Ring. We don't even need to answer the door anymore. We can just talk through our phone, to whoever's knocking on our door, right? Alexa now has a feature where it can tell when the delivery driver is delivering your parcel, and will give you a notification, like the delivery man just dropped your parcel. With Alexa, you can turn your lights in your home, without actually being in your home, right? So, there's so many things that's going to make our life more convenient, in ways that are so subtle that we won't even recognize. Like look at phones alone. Phones use a lot of AI on their camera, than you’d suspect. It's the reason why we don't have a 200 megapixel camera, because we don't need it. We can use AI to enhance the image for you. That's how Samsung Galaxy 22 Ultra, you could zoom into the moon, and it will give you a clear image of the moon, because it's using AI to enhance that image, and also references of what the moon looks like, to give you what the moon looks like. And also with, like, fingerprint as well. The more you use your fingerprint sensor on your phone, the better it is at unlocking your device when you touch the screen, again that's AI.

Let’s hypothesize. If you had complete control over how artificial intelligence is going to continue to develop, if it was just you. You get a say in how the future of artificial intelligence is going to continue, what would you find the most important thing to consider?

Probably with research. Research into space, or elements, or what physicists are doing, I think it would be much more useful for them to utilize it, because then they have something that essentially is connected to a whole lot of information. They can input their findings, and then they can get output of some research that's equally on their level. And with that, they can make some more discoveries with the help of AI, and hopefully with GPT 5 or 6, it can get to that level.

Yeah. Nice. So, it could be used as a very useful tool in fields of research, critical analysis, any of that kind of thing. Nice.

But, what is it that freaks you out – anyone, but you specifically right now - about the idea of putting a software like Chat-GPT into a robot, like into a physical robot? Because someone else told me about this as well, and I get it, like it's spooky. I wonder what it is specifically, that we're like, ‘it's so much more powerful’ now that it exists in our physical plane.

Well, I guess because then you give it sort of…

It's almost like giving it permission to exist, isn't it?

Yeah, it's right in front of you.

Can't ignore it now.

It’s not just the screen, I can't just turn off. Not only that, when it realizes what it can do with this new body, and how it can interact. It might learn a few things, or might come to a realization, that ‘now okay, maybe these guys are-‘, you know, ‘these guys are basically nothing to me, I'm just going to do what I want’. It's almost like it just going to a zoo and looking at all the animals.

Yeah.

Is there anything you want to promote? Any work that you're doing? Any, anything?

Not at the moment, but I will have something for artists to use, using AI. And hopefully I can show artists how to use it as a tool, rather than to be something that they should be scared of. Yeah, I’m just going to keep it like that for now, and when it's ready, it’s ready.

Awesome, brill. Thank you very much. That’s it!